iPTF13ehe is a hydrogen-poor superluminous supernova (SLSN) at z=0.3434, with a slow-evolving light curve and spectral features similar to SN2007bi. It rises within (83-148)days (rest-frame) to reach a peak bolometric luminosity of 1.3x$10^{44}$erg/s, then decays very slowly at 0.015mag. per day. The measured ejecta velocity is 13000km/s. The inferred explosion characteristics, such as the ejecta mass (67-220$M_odot$), the total radiative and kinetic energy ($10^{51}$ & 2x$10^{53}$erg respectively), is typical of a slow-evolving H-poor SLSN event. However, the late-time spectrum taken at +251days reveals a Balmer Halpha emission feature with broad and narrow components, which has never been detected before among other H-poor SLSNe. The broad component has a velocity width of ~4500km/s and has a ~300km/s blue-ward shift relative to the narrow component. We interpret this broad H$alpha$ emission with luminosity of $sim$2$times10^{41}$,erg,s$^{-1}$ as resulting from the interaction between the supernova ejecta and a discrete H-rich shell, located at a distance of $sim4times10^{16}$,cm from the explosion site. This ejecta-CSM interaction causes the rest-frame r-band LC to brighten at late times. The fact that the late-time spectra are not completely absorbed by the shock ionized CSM shell implies that its Thomson scattering optical depth is likely <1, thus setting upper limits on the CSM mass <30$M_odot$ and the volume number density <4x$10^8cm^{-3}$. Of the existing models, a Pulsational Pair Instability Supernova model can naturally explain the observed 30$M_odot$ H-shell, ejected from a progenitor star with an initial mass of (95-150)$M_odot$ about 40 years ago. We estimate that at least $sim$15% of all SLSNe-I may have late-time Balmer emission lines.